Blueprints for High Availability

"A resilient system is one that can take a hit to a critical component, and recover and come back for more in a known, bounded, and generally acceptable period of time."

"If you have success with a particular system configuration, reuse it everywhere!"

"Disk usage will always expand to consume all available resources."

"A collection of servers that failover among themselves must be thought of in a nontraditional way. The bottom-line component is no longer the server, but rather the application."

"Performing some thought experiments during this part of the design phase will prevent you from ending up in an unknown or unmanageable state later on."

"Understanding the different possible failures for a network of systems is critical for ensuring its availability."

"Database systems obey the ACID properties: Atomic, Consistent, Independent, and Durable."

"When replication is done properly you end up with two consistent and equally viable data sets, ideally in distinct physical locations, connected to distinct physical hosts."

"Investing in application recovery is akin to the message sent out by health management organizations: If you do a good job of early detection and prevention, you won't get too sick."

"Even if your building or your entire city is wiped out, your business can be restored on other computers from properly generated and stored backup tapes."

"If backup tapes can rebuild your business after a major outage, then they also can enable a competitor to steal your business when no outage has occurred. They must be treated at all times like the crown jewels of your company."

"As part of developing reliable and resilient systems, we strongly recommend building and maintaining relationships with your key vendor representatives."

"Disaster recovery is the hardest level to achieve because, almost by definition, you cannot really know what you are preparing for."